Answers are grounded in this paper and cite their sources.
Research on artificial intelligence–based computer-aided diagnosis and clinical decision support systems (AI-CDSS) is expanding, but integrated evidence on publication patterns, keyword trajectories, online search activity, and methodological characteristics remains limited. To characterize AI-CDSS research by integrating bibliometric analysis, Google Trends comparison, and structured content analysis of citation-ranked publications. Scholarly interest was assessed using bibliometric analysis of Web of Science records from 2016 through August 2026, including author-keyword co-occurrence, bibliographic coupling, and reference co-citation, whereas Google Trends was used to characterize online search activity. A structured content analysis of citation-ranked publications was conducted to examine applications, methodological characteristics, and reported limitations. We analyzed 19,809 included publications from 2016 through August 2026. Output rose approximately 21-fold, from 193 records in 2016 to 4,082 in 2025. Output was geographically concentrated within the included corpus, with a Gini coefficient of 0.869. Among records with identifiable corresponding-author country or territory, China, the United States, and India accounted for 61.33%. Scientific Reports was the most frequent source ( n = 916). Frequent artificial intelligence– and diagnosis-related keywords included machine learning ( n = 5,666), deep learning ( n = 4,022), and computer-aided diagnosis ( n = 1,512). Bibliographic coupling identified 15 publication communities, and reference co-citation identified seven reference communities. In the monthly level-series comparison, 9 of 10 keyword concepts exhibited significant associations with Google Trends relative search interest, but none of the 10 first-difference associations remained significant after Bonferroni adjustment. Content analysis of 215 citation-ranked publications, including 191 empirical publications, showed frequent use of convolutional neural networks in diagnostic and radiology applications. Common limitations included limited external validation, infrequent calibration reporting, and sparse workflow-level or impact evaluation. AI-CDSS research has rapidly expanded around machine learning, deep learning, and diagnostic applications, but the analyzed citation-ranked publications provide more evidence of methodological development than of clinical deployment. Integrating bibliometric, search-activity, and content-based perspectives shows limited short-term correspondence between online search activity and scholarly keyword use and highlights the need to move beyond model development toward independent external validation, prospective workflow and patient-outcome evaluation, calibration, fairness, and postdeployment monitoring.
No takes yet. Share an insight, caveat, or question.
Chien et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: